This paper reports on a study of the interactions between text chat and audio modalities in L2 interaction in a synthetic (virtual) world and observes whether the text chat modality was used for corrective feedback and the characteristics of the latter. This is examined within the context of a hybrid Content and Language Integrated Learning design workshop. This course involved 17 students of architecture whose L2 was either French or English and for which the synthetic world environment Second Life was employed for distance language sessions. Using multimodal transcriptions of the interaction data from these sessions, it was found that text chat was employed for content-based interaction concerning the task as well as feedback concerning non-target-like errors in the audio modality. Feedback predominantly concerned lexical errors and was offered in the form of recasts. The multimodality of the environment did not appear to cognitively overload students who frequently responded in the audio modality to corrective feedback offered in the text chat. The study highlights the need to train language tutors who wish to exploit synthetic worlds to use the text chat in parallel with the audio to support learners' verbal production with respect to verbal participation and proficiency.